Muse Glimmer 30B on QuickSilver Pro
Muse Glimmer 30B is Meta's dense 30B open-weight multimodal model, distilled from Muse Spark and tuned for autonomous agents — long-horizon tool use with extended reasoning over a 131K-token context. On QuickSilver Pro it is $0.30 input / $1.20 output per million tokens, at parity with the OpenRouter endpoint selected by this route, with no markup or discount.
At a glance
Budget long-horizon agents and everyday coding with Muse Spark-style reasoning at a fraction of the price, over a 131K-token context.
Pricing comparison ($/1M tokens)
| Provider | Input | Output | vs QSP |
|---|---|---|---|
| QuickSilver Pro | $0.30 | $1.20 | — |
| OpenRouter (meta/muse-glimmer-30b) | $0.30 | $1.20 | same |
When to use
Reach for Muse Glimmer 30B when you want Muse Spark's agentic style at a small-model price: everyday coding loops, tool-calling agents that plan and iterate across many steps, and batch workloads where token spend dominates. It accepts images alongside text, so an agent can iterate against screenshots and rendered output, and its 131K-token context covers a substantial working set.
When to use something else
For repository-scale working sets it tops out at 131K tokens — step up to Muse Spark 1.2 ($1.00/$3.40, 1M context) or Kimi K3 when the whole repo must stay in context. Reasoning is mandatory here, so it spends output tokens even on trivial turns; rein that in with `reasoning_effort: "low"` where cost matters, or use DeepSeek V4 Flash ($0.06/$0.18) for high-volume single-shot calls that don't need thinking at all.
Quickstart (curl)
curl https://api.quicksilverpro.io/v1/chat/completions \
-H "Authorization: Bearer $QSP_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "muse-glimmer-30b",
"messages": [{"role": "user", "content": "Hello!"}]
}'OpenAI-compatible. One-line migration via base_url.
FAQ
Glimmer is the small, open-weight sibling: a dense 30B model distilled from Muse Spark, tuned for autonomous agents, at roughly a third of Spark's per-token price with a 131K-token context versus Spark's 1M. Spark remains the pick for repository-scale coding and the hardest long-horizon work; Glimmer covers everyday agent loops where cost and iteration speed matter more than peak capability. Both take text and images as input and reason before answering.
Reasoning is always on — like Muse Spark, it is mandatory on this endpoint and cannot be disabled. Use `reasoning_effort` (`low`, `medium`, `high`, or `xhigh`; the default is `medium`) to trade depth against latency and cost. Reasoning tokens are drawn from your `max_tokens` budget before the visible answer, so set `max_tokens` generously — 1024 or more — or the reply can come back empty because the budget was spent thinking.
$0.30 input / $1.20 output and $0.04 cache read per million tokens, matching the route-selected OpenRouter endpoint (deepinfra bf16) with no markup or discount. The route uses its pinned provider order or the cheapest eligible endpoint under sort: price and its quantization filters; the /models headline can differ. More expensive fallback hosts do not change your price. Use base_url=https://api.quicksilverpro.io/v1 and model="muse-glimmer-30b".
Yes — it's an OpenAI-compatible chat completions endpoint, with the Responses API available too. Streaming, tool calling, and json_schema strict mode all work, and it accepts image input. Set base_url=https://api.quicksilverpro.io/v1 and model="muse-glimmer-30b" in any OpenAI-SDK agent framework.